[OSM-talk] Mechanical Edit?

stevea steveaOSM at softworkers.com
Wed Jul 28 06:48:58 UTC 2021

Note there are some repeated themes about this relatively-advanced OSM topic:

There are "tricks" (like finding mass edits without certain particular issues), "difficulty" (in being certain that crap data ARE crap data while others simultaneously and earnestly say "...but, I'm improving the map!"), good intentions (usually), and that beginning editors / new OSM users should STAY AWAY from ANYthing "mechanical" or "automated."  Every single one of these rings true for me going back to my Day 1 of joining OSM and in dealing with a lot of absolute junk edits (largely by others, but I goofed up a bit in my early days too, that is part of how we learn — and I DID correct those mistakes).  Even today, nobody is perfect, we are all human and make mistakes and they do put erasers on the ends of pencils.  OSM must have methods for people to make (minor, hopefully) mistakes and learn from them.  Mechanical edits are fraught, rife with the possibility of mistakes (though again, ARE NOT ALWAYS WRONG) and so I'll emphasize "don't learn doing this, do so after becoming a learned and practiced editor."

This is not easy, either talking about it, doing it or suggesting how others do it.  There really are a LOT of inputs into the equation of "should I mechanically edit?"  Even if (again), it is simply Validator making warning-suggestions, we DO want to improve these (if we can), but we also say "if you can't, simply ignore these."  Validator is on the right path as it suggests that, we might build upon that idea.

If you are a super-experienced editor who has earned the respect of your local community (and better, even wider than that), you might think about it, you might even actually do it.  But "take a deep breath" first, think hard before you do and I'm only talking to a small fraction of you anyway as I say this; you likely already know this.  So, for the rest of you, get some experience and earn some respect and then cross the bridge to considering how a lightweight "mechanical" (flavored) edit might improve the map.  It might.  It might not.  (See above about "tricky" and "difficult" and "crap data").

> On Jul 27, 2021, at 11:30 PM, Mateusz Konieczny via talk <talk at openstreetmap.org> wrote:
> Jul 28, 2021, 00:17 by frederik at remote.org:
> Hi,
> On 7/27/21 16:50, Mateusz Konieczny via talk wrote:
> Changesets should be assessed on the *quality* not quantity.
> for automated/semiautomated edits quality includes also quality of
> documentation/discussion/consensus.
> Yes, we have definitely reverted undiscussed map edits where 99% of edits were an improvement and 1% were making things worse, just to prove a point.
> Also, with automation: editing 100 000 objects may be far easier than cleanup 1000 damaged ones.
> So sometimes the mass replace part is just 2% of work and whoever made it expects
> 98% of work of cleanup to be done by others.
> (the trick is finding mass edits without that issue and discussing before edit to confirm
> that with other. BTW, making local edits is often easier as many local communities have
> less of "I oppose all automated edits because it will modify objects"/
> "I am against any automation" responses than global ones)

More information about the talk mailing list